FiberDefect-CT dataset for "An efficient segmentation framework for yarns and defects in woven composites based on structure-aware and generative augmentation" (Composites Communications)

FiberDefect-CT This dataset contains micro-CT slice images and corresponding pixel-level annotations for woven composite materials, collected to support the study reported in “An efficient segmentation framework for yarns and defects in woven composites based on structure-aware and generative augmentation”, published in Composites Communications. The dataset includes partial-slice CT images of nine different materials acquired in two directions, together with corresponding pixel-wise annotations of warp yarns, weft yarns, and pores. It is intended to support research on semantic segmentation, defect characterization, and structural analysis of woven composites. How to cite If you find the FiberDefect-CT dataset useful in your research, please cite our work. Ying Cao, Ziwei Nie, Fangfang Sun, Qian Zhao, Chao Li, Xiaoping Yang, An efficient segmentation framework for yarns and defects in woven composites based on structure-aware and generative augmentation, Composites Communications, Volume 64, 2026, 102800, ISSN 2452-2139, https://doi.org/10.1016/j.coco.2026.102800. Email: yingcao@smail.nju.edu.cn A GitHub repository is linked to this dataset (https://github.com/YingCao0305/FiberDefect-CT).